Transcription
Everyone is talking about Israel launching a supercomputer. Everyone is looking at a thousand of the latest Nvidia accelerators. No one is asking the main question: where did they come from, when the whole world is in line? January 19, 2026, Tel Aviv, Innovation Authority press conference. Journalists photograph the minister, headlines write themselves: "Israel - leader in artificial intelligence, 315 million dollars into the future. The first national supercomputer for startups and science. A celebration, pride, a technological triumph. But here's what didn't make the news. 3 months before the official launch, the company Nebius quietly won the tender. Not an Israeli tech giant. Not an American corporation. Nebius is the brainchild of Arkady Volozh, the man who created Yandex, Russia's largest internet search engine, the man who left his empire, refusing to serve Putin. A man who became an Israeli citizen in 2017, 5 years before the invasion of Ukraine. The official version: the best offer was chosen. Nebius offered four times more computing power than required in the basic tender. 1 billion 300 million shekels, 140 million dollars. A thousand Nvidia B200 Blackwell accelerators - chips that just a month ago were the most scarce on the planet. Google is waiting in line for these chips. Meta is planning purchases 2 years in advance. China is trying to circumvent the American embargo and create analogues. And Israel received a thousand pieces right now. Three questions that hung in the air after that press conference. First. How did a country with a population of 10 million people overtake tech giants in the global queue for the planet's most scarce chips? NVIDIA produces B200 Blackwell in limited batches. Demand exceeds supply by tens of times. Priority is given to the largest clients: Amazon, Microsoft, OpenAI. But Israel receives 1,000 accelerators at a time when even Chinese companies with billion-dollar budgets cannot buy a single one. Second. Why did critical national infrastructure, a supercomputer on which defense systems will be trained, end up in the hands of a company founded by an immigrant from Russia? Nebius. An unfamiliar name. Headquarters in Amsterdam, but behind it stands Arkady Volozh. A man who managed a tech empire for 25 years in a country where the internet is controlled by the Kremlin. Trust or risk? Third. Why did the national artificial intelligence program, announced several years ago, stall for so long that it had to be saved by a single deal with a single company? Israel is the fourth largest AI innovation power in the world. 47% of all investments in Israeli startups go to AI-related companies. But until January 19, 2026, the country did not have a single national supercomputer for training models. Today, I will show you how the global war for AI chips has turned technological scarcity into a geopolitical weapon. I will reveal why the choice of a Russian entrepreneur for critical infrastructure is not a mistake, but the only possible solution. And I will explain how small countries are rewriting the rules of the big game not by scale, but by speed. And perhaps after this video, you will understand, a supercomputer is not just hardware in a data center. It is a matter of survival, a matter of independence. A question of whether Israel will be a technological power or a technological colony in 5 years. But you think that's all? Here's what they didn't say at the press conference. Until January 19, 2026, Israeli startups, the very ones that create security, defense, and cyberattack systems, trained their artificial intelligence models on servers from Amazon Web Services, Google Cloud, Microsoft Azure. Read carefully. Algorithms that protect Israel's borders were processed on infrastructure belonging to American corporations. Imagine the scene. The office of an Israeli startup developing a threat detection system for the army. Engineer Noam Levy uploads data for model training: terabytes of information, satellite images, drone footage, movement patterns. Everything is uploaded to Google servers in Virginia, or Ireland, or Singapore. Noam doesn't know exactly, the cloud is distributed. He only knows that if Google decides to restrict access tomorrow, the project will stop. If the American government demands an audit of the data, secrets will cease to be secrets. This is not paranoia, this is the reality of technological dependence. In 2023, Google restricted access to its GPUs for companies from several Middle Eastern countries under the pretext of complying with export controls. In 2024, Amazon introduced quotas for using computing power for training large models. Priority is given to its own projects. Israel was not subject to direct restrictions, but the pattern is obvious. Whoever controls the hardware controls the rules. Now look at the numbers. 47% of investments in Israel go to AI companies. This is the highest share in the world after the United States. 2,000 startups work with artificial intelligence. High-tech exports - 50 billion dollars annually. Half are related to AI technologies. But until January 19, 2026, this entire ecosystem stood on a foreign foundation. Do you know what a GPU is in today's world? It's not just chips for graphics cards. It's the new oil. Countries that control the production and distribution of graphics accelerators dictate who will be the leader in artificial intelligence for the next 10 years. The US produces through Nvidia and AMD. China is trying to create analogues but is 3-4 years behind due to American sanctions. Europe imports, the Middle East imports. Israel, until January 19, imported access to clouds, not having a single national-scale cluster of its own. Here's a comparison that paints a picture. 1,000 Nvidia B200 Blackwell accelerators. Do you know how much they cost on the open market? $30-40,000 for one chip. Multiply by 1,000. $40 million just for the hardware. Without infrastructure, without cooling, without electricity. But that's not the main thing. The main thing is, they cannot be bought. They are distributed by priority. Google orders tens of thousands, receives them a year later. Meta builds data centers with hundreds of thousands of GPUs, a 2-year queue, and Israel receives 1,000 pieces in a few months. Ask yourself the question: "How? How did a small country bypass tech giants?" The answer is not in money. $315 million is the budget for one Meta project. The answer is in connections, in knowing how the global supply chain works, in the ability to negotiate with those who control scarcity. Imagine an alternative scenario. 2026. The conflict in the Middle East escalates. Israeli defense companies urgently need to update AI systems for threat analysis. They need computing power, terabytes of data for model training, but access to clouds is limited. Amazon introduces quotas. Google demands an audit. Microsoft slows down processing due to technical problems. The official version: overload. The reality: political pressure. In this scenario, Israeli startups stop. Updates to critical systems are delayed for weeks. Enemies gain a temporary advantage. The price of technological dependence is measured not in dollars, it is measured in lives. But Israel chose a different path. January 19, 2026, the day dependence ended. Now Israeli companies train models on their own infrastructure. Defense systems are processed on domestic servers. Data does not leave the country. Access is controlled by the state, not by Silicon Valley corporations. But here's the question that hung in the air after the announcement. Where did these thousands of chips come from? Why Nebius? Who is Arkady Volozh? And why did his company get the contract for critical national infrastructure? The answer leads us to a man who knows the price of technological freedom better than anyone, because he once lost an empire by refusing to serve a regime. And now he is building the future for another country, one that chose freedom over control. February 2022, Moscow. The third week after the invasion of Ukraine. The office of Yandex, Russia's largest tech company. Valuation $20 billion. The board of directors convenes for an emergency meeting. On the agenda, one question. What to do with the founder? Arkady Volozh sits at the negotiating table. For 25 years, he built this empire. A search engine that processes 63% of queries in Russia. Maps, taxi, food, music, cloud services. 19,000 employees, billions in revenue. But in 3 days, the European Union will add him to the sanctions list for helping the Russian regime control information. He looks at his colleagues. To stay means to finally submit to the Kremlin. Yandex has been under pressure for a year. Censorship of search results, transfer of data to special services, blocking of opposition websites. If he stays, he will become part of the war machine. If he leaves, he will lose everything. 25 years of work. The company he built from scratch, status, power, billions. Arkady stands up. I am leaving immediately. The board of directors is silent. He leaves the office, never to return. Two months later, he will give up all his Yandex shares completely, without compensation. Europe will lift sanctions, but the empire is lost forever. This is what they don't write in official biographies. Volozh became an Israeli citizen in 2017, 5 years before the war. Not an escape, a strategy. He understood that Russia was moving towards autocracy, that the internet was becoming a tool of control. Technological business is impossible without compromises with power. Israeli citizenship was plan B, a fallback option, a guarantee of freedom. But here's what makes this story critical to our investigation. Volozh is not just an immigrant, he is a man who understands the logic of technological confrontation from the inside. He saw how the Kremlin demanded access to user data, knew how special services pressured companies, experienced firsthand what it means to build a business in conditions where the state controls infrastructure. When he finally left Yandex in 2022, the tech world asked, what's next? A man with experience managing a tech empire, with an understanding of how big data, cloud services, and computing clusters work, with a reputation as someone who chose principles over billions, where will he direct this experience? The answer came a year later. 2023. Volozh announces the creation of Nebius. Not a search engine, not a taxi. Cloud infrastructure, data centers, computing power for training artificial intelligence models. Exactly what the world is experiencing a severe shortage of. September 2025. The Tel Aviv Innovation Authority announces a tender for the creation of a national supercomputer for artificial intelligence. Requirements: at least 250 of the latest GPUs. Budget: 100 million dollars. Deadline: launch by the end of 2025, early 2026. Three major companies submit applications. An American tech giant offers 300 GPUs - a standard, time-tested solution. A European consortium promises 400 accelerators, but with delivery in a year and a half. Nebius. Arkady Volozh comes with an offer that seems impossible. 1,000 NVIDIA B200 Blackwell accelerators, four times the minimum requirements. Launch by January 2026. Members of the tender committee look at the numbers: 1000 B200. These are chips that Nvidia only started producing in mid-2025. Google is in line. Meta plans purchases for 2 years in advance. OpenAI received its first batch through Sam Altman's personal connections. How can a company that is only 2 years old promise to deliver a thousand of the most scarce chips on the planet in 3 months? One of the committee members asks the question. Mr. Volozh, how do you guarantee delivery? Do you have a contract with Nvidia? Arkady opens a document. Yes, the contract is signed. Nebius is one of Nvidia's strategic partners in Europe. Why? Because in 2 years, Volozh has built not just a company, he has built a network. This is what makes Volozh's story key to understanding the entire project. After leaving Yandex, he didn't start from scratch. 25 years of managing a tech empire gave him what startups lack: connections, trust, the reputation of a man who keeps his promises. NVIDIA knows him. Google worked with Yandex. Amazon saw how Yandex built cloud services like Yandex Cloud. When the world faced a GPU shortage in 2023-2024, major players began building long-term partnerships with manufacturers. Not one-off purchases, but strategic alliances. Nvidia chose partners not by the size of the check, but by their ability to scale infrastructure. Nebius made it onto this list not because it's big, but because behind it stands a man who has already proven he knows how to build big. Now, a question for you. Imagine you are a member of the tender committee. You have three offers before you. An American company is reliable, but 300 GPUs. A European consortium - 400 GPUs, but in a year and a half. Nebius - 1,000 GPUs in 3 months. But the company is young, the founder is an immigrant from Russia. What would you choose? The safety of the proven or the ambition of someone who promises the impossible? Israel chose ambition, because behind the ambition stood a man with a track record, because the alternative of waiting a year and a half meant a year and a half of dependence on foreign clouds, because 300 GPUs is a compromise, not a solution. But here's the irony that escapes a superficial glance. Critics said, how can you trust critical infrastructure to a man who worked in Russia for 25 years? But it was precisely his experience working in Russia that made Volozh the ideal choice. He knows what it's like to build technology in conditions where the state can intervene at any moment. He knows how to protect data, he knows the price of independence. Remember the scene from the beginning. February 2022. Volozh leaves the Yandex office, loses billions, but retains his freedom. This choice is not weakness. It is understanding. In a world where technology determines power, freedom is more valuable than money. Israel bet on a man who proved this at personal cost. December 2025, a site near Tel Aviv. Builders are installing the last server racks. Nvidia trucks are delivering boxes with B200 Blackwell. Thousands of accelerators, each the size of a book. Each capable of processing trillions of operations per second. All together, 16,000 petaflops of computing power. Volozh stands in the center of the data center. Engineers are checking the cooling system. B200s generate a huge amount of heat. Electricians are testing power lines. Consumption like a small city. In a month, this complex will become the heart of the Israeli AI industry. 70% of the capacity will go to startups at preferential prices, 30% to academic researchers. Everything is under state control. One of the engineers approaches Volozh. We are ready. The system launches on January 19. Volozh nods. 25 years ago, he built a search engine for Russia. Today, he is building computing infrastructure for Israel. One story ended with giving up an empire, another begins with creating a foundation for the future of an entire country. What if Israel had chosen differently? Imagine an alternative scenario. The tender committee decides: Nebius is too risky. They choose an American company, 300 GPUs, a proven solution, or a European consortium, 400 GPUs in a year and a half. In the first case, the 300 accelerators are divided among hundreds of startups and researchers. Queues and limitations. Each project gets access for a few hours a week instead of around the clock. Training large models takes months. Israeli companies continue to use Amazon and Google because national infrastructure is insufficient. In the second case, a year and a half of waiting. That's 18 months during which competitors from the US and China overtake Israel. 18 months of dependence on foreign clouds, 18 months when defense systems are processed on servers to which foreign governments theoretically have access. But Israel chose the third path, bet on a man who promised the impossible and delivered. Three months after signing the contract, 1,000 B200 Blackwells are in a data center near Tel Aviv. Not 300, not in a year and a half. A thousand, here and now. Ask yourself: why was Volozh able to do this? The answer lies not in money or technology. It lies in understanding how the global supply chain works in conditions of scarcity. NVIDIA produces a limited number of B200s. Demand is huge. Distribution is not based on who pays more, but on who is a strategic partner. In 2 years, Volozh made Nebius a strategic partner, not with promises, but with real projects. Data centers in Europe, cloud services for business, infrastructure that grows and scales. NVIDIA sees that Nebius is not a speculator who will buy chips and resell them at a higher price. It is an infrastructure builder, a long-term player. When Israel announced the tender, Nebius used these relationships not for resale, but for a project that would become a showcase. A national supercomputer, access for startups and science, computing power for defense systems. NVIDIA agreed because it's not just a deal, it's a demonstration of technology. A showcase of how a small country can become a leader through the proper use of advanced chips. Here's the connection that no one is explicitly stating. Volozh's story is not just a personal triumph, it is a symptom of a deeper process. The world has entered an era of technological sovereignty. Countries are divided into those who produce chips and those who depend on supplies. But there is a third category. Those who do not produce, but know how to build relationships to gain access at a critical moment. Do you see the connection? Topic A. Chips as weapons. GPU scarcity turns small countries into hostages. Topic B. A man who knows how to work under conditions of control and scarcity because he spent 25 years in Russia. Now you have two facts. Problem. Israel depended on foreign clouds. Solution. Volozh used connections and reputation to get a thousand scarce chips in 3 months. But the main question remains: are a thousand B200s enough when Google builds clusters of 100,000 GPUs, when Meta invests $30 billion in its own AI infrastructure, when China, despite lagging behind, produces millions of less powerful chips and wins by quantity? Israel with a population of 10 million against tech empires with trillion-dollar budgets. How to compete at all? The answer overturns the entire understanding of what it means to be a small country in the era of artificial intelligence. Because Israel is not trying to compete by scale. It is rewriting the rules of the game. And here's how it works: when speed is more important than size, and smart distribution wins over brute force. This is where the real story begins. Not about chips, but about strategy. Not about technology, but about survival through innovation. So, Israel, a country without its own computing power, received a thousand of the most scarce chips on the planet in 3 months. Volozh, a man who lost an empire in Russia, built infrastructure for his new homeland. Do you see the pattern? It's not about money, not about size. It's about how small countries survive when technology becomes a weapon. This is what it means. January 20, 2026. Exactly one day after the official launch of the supercomputer, Israeli startup Perception AI gains access to the first computing resources. A company of eight people. An office in Tel Aviv the size of a three-room apartment. They are developing a satellite image analysis system for real-time threat detection. The customer is the Israel Defense Forces. The startup's founder, Shira Rosenfeld, 32, a former officer of Unit 8200. Until January 20, Shira rented computing power from Amazon Web Services. The cost was $15,000 per month for GPU access. Training one model took 2 weeks. Each iteration another 2 weeks. Each algorithm improvement - $30,000 and a month of waiting. Now she connects to the national supercomputer. The cost is 7,000 shekels per month, less than $2,000. A preferential rate for defense projects. Model training speed is 3 days instead of 2 weeks. Data does not leave Israel. Shira launches the first calculation. After 72 hours, she gets a result that used to take a month. This is what changes. Not abstractly, but concretely. Perception AI will save $160,000 per year on computing power, accelerate development fivefold, and create six new jobs for engineers. And most importantly, the security system they are creating will be processed on infrastructure controlled by the State of Israel, not a corporation from Seattle. Multiply Shira by 200. That's how many startups and research projects will get access to the national supercomputer in the first year. 70% of the capacity is reserved for the private sector, 30% for academic research. This is not just infrastructure, it is an innovation accelerator. This is how the model works. Traditional approach: each company buys or rents its own GPUs. Google owns hundreds of thousands of chips solely for internal projects. Meta builds data centers, closed to external users. The result: concentration of power in the hands of giants - scarcity for everyone else. This is the logic of scale. Whoever is bigger is stronger. The Israeli model overturns this logic. 1000 B200 Blackwells are not for one company, but for an ecosystem. Each accelerator works around the clock on dozens of projects simultaneously. A startup uses the capacity for 3 days, frees it up for the next. A university trains a model for a week, transfers resources to another lab. One chip generates value for 20 projects a year instead of one. This is the survival formula for small countries. Not to produce more, but to use more efficiently. Not to own everything, but to distribute more intelligently. Not to compete by scale, but to win by speed of circulation. Now look at the pattern. Israel is not the only small country that is betting on artificial intelligence. Singapore is investing a billion dollars in a national AI program. Switzerland is building alliances with tech giants for access to computing resources. South Korea is subsidizing startups working with machine learning. What do these countries have in common? They are not trying to become the new China or America. They are not building chip factories for hundreds of billions. They are creating ecosystems of maximum efficiency. Singapore. 6 million population, but ranks in the top 10 for AI patents per capita. Switzerland 8 million, but leads in AI integration into industry. Israel - 10 million, but is fourth in the world for AI innovations. The pattern is obvious. In the era of technological empires, small countries survive through intensity, not extensiveness, through speed of decision-making, through lack of bureaucracy, through direct connection between the state, science, and business. Here's a concrete example of contrast. The European Union, 450 million population, announced a plan to create sovereign AI infrastructure in 2023. Budget: 10 billion euros. 3 years later, disputes between countries over data center locations, debates about regulation, agreement on standards. Result: zero working national supercomputers for the civilian AI sector. Israel, 10 million population, announced a tender in September 2025. Four months later, the supercomputer is launched. Startups get access the next day. Not 10 billion, but 315 million. Not 3 years of bureaucracy, but 4 months from decision to result. Do you see the advantage? Small countries lose in scale, but win in speed. Imagine the scene. July 2027, a year and a half after the launch of the Israeli supercomputer. Brussels. An artificial intelligence conference. On stage, the Israeli Minister of Innovation. He shows statistics. In a year and a half, the national supercomputer has processed 240 projects. 30 of them have received commercial funding, eight have entered the international market, three have signed contracts with the armies of other countries. In the audience are representatives of European governments. Their sovereign AI program is still in the budget negotiation stage. A minister from Germany asks a question: "How did you achieve such speed of implementation?" The Israeli colleague replies: "We didn't try to build the biggest. We built the most efficient and opened access to those who create real value: startups and scientists." This is not fantasy, it is an extrapolation of an existing trend. Israel historically wins not by the size of its army, but by the quality of training, not by the number of universities, but by the intensity of research, not by the scale of its economy but by its high-tech GDP. The logic is the same. The national supercomputer is not an attempt to catch up with Google in terms of GPU count. It is a bet that 1,000 chips in the hands of 200 innovative projects will create more value than 100,000 chips in the closed data center of one corporation. Now let's go back to the beginning. Remember Noam from the first part, the startup engineer who uploaded defense system data to Google servers in Virginia. January 20, 2026, the day his life changed. Noam connects to the national supercomputer. First launch. The model begins training on Israeli infrastructure. Data is protected. Speed is high, cost is seven times lower than Amazon. He looks at the screen. For the first time in 5 years of working on the project, he is not thinking: "What if access is blocked?" For the first time, he knows: the infrastructure belongs to his country. Control is in the hands of the state, which is interested in the success of his project. The decision made in September 2025 to choose Nebius, to bet on 1000 B200s, to launch in 4 months. This decision has changed the lives of thousands of engineers, scientists, and entrepreneurs. Shira from Perception AI, Noam from the defense startup, hundreds of others whose names we don't know, but whose contribution will determine Israel's technological future. This is the current reality that makes this story critically important today. January 20, 2026, is not just the date of the supercomputer launch. It is a symbolic moment in the global war for technological sovereignty. Look at the context. December 2025. The United States is tightening export controls on advanced chips. China can no longer import anything more powerful than a certain level. Europe has fallen into a gray zone, not under sanctions, but at the end of the supply queue. The Middle East is considered a potential zone for technology leakage to China and Russia. In these conditions, Israel receives 1,000 B200s. Not because it circumvented sanctions, but because it proved it is a strategic ally of the US, a democratic country, a partner in confronting authoritarian regimes. Nvidia does not sell chips just like that. Every shipment to a region with instability is a political decision coordinated with Washington. The choice in favor of Israel is a message. Technological democracies get access, others do not. The supercomputer in Tel Aviv is not just infrastructure for startups. It is a geopolitical statement about which side Israel is on in the global confrontation. Ask yourself: what will happen in 5 years? 2031. The world is divided into technological blocs. The US controls the production of advanced chips. China has created its own ecosystem, less powerful but vast in scale. Europe hesitates between sovereignty and dependence on American corporations. And small countries, those that have not invested in their own infrastructure, have turned into technological colonies. Their startups operate on foreign servers. Their data is processed abroad. Their innovations depend on the mercy of these giants. They import not only technology, they import decisions about what they are allowed to create. Israel chose a different path. January 19, 2026. A point of bifurcation. The day the country said: "We will not be a colony. We will build our own foundation, not as large as an empire's, but sufficient to control our future." In 5 years, this supercomputer will become obsolete. B200 Blackwells will be replaced by B300s or B400s. A thousand accelerators will seem like a ridiculous number compared to clusters of millions of GPUs. But the principle will remain. The principle that technological independence is not a luxury, it is a condition for survival. That investment in one's own infrastructure is as much a part of defense as tanks and airplanes. This is the lesson Israel is teaching the world through this story. You cannot defeat empires with their own weapons. China is trying to catch up with the US in chip production, spending a trillion dollars, and is years behind. Europe is trying to create its own tech giants, subsidizing startups, losing to Silicon Valley. Small countries cannot compete by scale, but they can rewrite the rules of the game, create a model where every dollar, every chip, every talent is used with maximum efficiency, where the state, business, and science work as a single system, where decisions are made in months, not years. Israel has proven this time and again. In agriculture, drip irrigation versus large-scale field flooding. In the army, small, high-tech versus large, poorly trained. In the economy, high-tech instead of natural resources. Now in artificial intelligence, smart distribution instead of brute force of quantity. January 19, 2026. This day will go down in textbooks not as the date of the supercomputer launch, but as the moment when a small country showed that in the era of technological empires, it is possible to survive without submitting, without becoming a colony, but by building its own model, compact, efficient, sovereign. Volozh's story, Shira's story, Noam's story. These are not separate plots. They are parts of one big story about how those who refuse to be weak survive. Volozh lost an empire, built infrastructure for his new homeland. Shira depended on Amazon, gained access to national resources. Noam feared blocking, gained control. Each of them is a symbol of choice. The choice not to accept the rules imposed on you. The choice to build your own path, even if it seems impossible. The choice to believe that size does not determine strength, but the ability to use what you have. This is a story about Israel. A country that in its 78 years of existence has proven that being surrounded by enemies is not a sentence, lack of resources is not weakness, small size is not a disadvantage. It is an incentive to be smarter, faster, more efficient. It is a reason not to compete with empires, but to rewrite the rules so that empires envy. A thousand Nvidia B200 Blackwell accelerators, 315 million dollars, 4 months from decision to launch. Not the largest supercomputer in the world, but the first in the world where 70% of capacity is open to startups at preferential prices, where each chip works for dozens of projects, where technological sovereignty is not a slogan, but a reality. This is what it means to be strong. Not when you have the most, but when what you have is used so effectively that it creates an advantage against those who have 10 times more resources. This is not about technology, it is about choice. The choice to never give up. The choice to turn every crisis into an opportunity. The choice to build the future, even when everyone says it's impossible. January 19, 2026, the day Israel made this choice again and showed the whole world: small countries are not doomed to be colonies. They can be leaders. If they choose speed over scale, efficiency over quantity, sovereignty over dependence. This is our story. The story of a people who survive not because they are the strongest, but because they are smarter, faster, more persistent, and never stop believing. The future is built not by those who have more resources, but by those who use them best.